We introduce a general framework for modeling functionally diverse problem-solving agents. In this framework, problem-solving agents possess representations of problems and algorithms that they use to locate solutions. We use this framework to establish a result relevant to group composition. We find that when selecting a problem-solving team from a diverse population of intelligent agents, a team of randomly selected agents outperforms a team comprised of the best-performing agents. This result relies on the intuition that, as the initial pool of problem solvers becomes large, the best-performing agents necessarily become similar in the space of problem solvers. Their relatively greater ability is more than offset by their lack of problem-solving diversity.A diverse society creates problems and opportunities. In the past, much of the public interest in diversity has focused on issues of fairness and representation. More recently, however, there has been a rising interest in the benefits of diversity. In the legal cases surrounding the University of Michigan's admissions policies and in efforts to curtail affirmative action in California, Texas, and elsewhere, there have been claims that diverse perspectives improve collective understanding and collective problem solving. Coincident with this political and legal wrangling has been an effort on the part of scholars to identify how to exploit this diversity both in solving hard computational problems (1, 2) and in human organizations (3).In the common understanding, diversity in a group of people refers to differences in their demographic characteristics, cultural identities and ethnicity, and training and expertise. Advocates of diversity in problem-solving groups claim a linkage among these sorts of diversity (which we will refer to as identity diversity) and what we might call functional diversity, differences in how people represent problems and how they go about solving them. Given that linkage, they conclude that, because of their greater functional diversity, identity-diverse groups can outperform homogeneous groups (4-6).Building on earlier ideas from the psychology and artificial intelligence literatures (7), we describe a mathematical framework for modeling problem solvers that captures the functional diversity that cognitive psychologists and organizational theorists claim is correlated with identity diversity. In our framework, agents possess internal representations of problems, which we call perspectives, and algorithms that they use to locate solutions, which we call heuristics. Together, a perspective-heuristic pair creates a mapping from the space of possible solutions to itself. A diverse group is one whose agents' mappings are diverse. Our perspective-heuristic framework is not minimal, because we show in an earlier paper (8) that two problem solvers with distinct perspectives and heuristics can act identically in the space of solutions. However, the advantage of the full framework is that it generalizes models in the computer science literature...

No abstract

No abstract

No abstract

Each year, corporations spend billions of dollars on diversity training, education, and outreach. In this article, I explain why these efforts make good business sense and why organizations with diverse employees often perform best. I do this by describing a logic of diversity that relies on simple frameworks. Within these frameworks, I demonstrate how collections of individuals with diverse tools can outperform collections of high "ability" individuals at problem solving and predictive tasks. In problem solving, these benefits come not through portfolio effects but from superadditivity: Combinations of tools can be more powerful than the tools themselves. In predictive tasks, diversity in predictive models reduces collective error. It's a mathematical fact that diversity matters just as much as highly accurate models when making collective predictions. This logic of diversity provides a foundation on which to construct practices that leverage differences to improve performance.

In this paper, we identify two distinct notions of accuracy of land-use models and highlight a tension between them. A model can have predictive accuracy: its predicted land-use pattern can be highly correlated with the actual land-use pattern. A model can also have process accuracy: the process by which locations or land-use patterns are determined can be consistent with real world processes. To balance these two potentially conflicting motivations, we introduce the concept of the invariant region, i.e., the area where land-use type is almost certain, and thus path independent; and the variant region, i.e., the area where land use depends on a particular series of events, and is thus path dependent. We demonstrate our methods using an agent-based land-use model and using multitemporal land-use data collected for Washtenaw County, Michigan, USA. The results indicate that, using the methods we describe, researchers can improve their ability to communicate how well their model performs, the situations or instances in which it does not perform well, and the cases in which it is relatively unlikely to predict well because of either path dependence or stochastic uncertainty.

We develop a model of two-party spatial elections that departs from the standard model in three respects: parties' information about voters' preferences is limited to polls; parties can be either office-seeking or ideological; and parties are not perfect optimizers, that is, they are modelled as boundedly rational adaptive actors. We employ computer search algorithms to model the adaptive behavior of parties and show that three distinct search algorithms lead to similar results. Our findings suggest that convergence in spatial voting models is robust to variations in the intelligence of parties. We also find that an adaptive party in a complex issue space may not be able to defeat a well-positioned incumbent.

Public health research and practice have been guided by a cognitive, rational paradigm where inputs produce linear, predictable changes in outputs. However, the conceptual and statistical assumptions underlying this paradigm may be flawed. In particular, this perspective does not adequately account for nonlinear and quantum influences on human behavior. We propose that health behavior change is better understood through the lens of chaos theory and complex adaptive systems. Key relevant principles include that behavior change (1) is often a quantum event; (2) can resemble a chaotic process that is sensitive to initial conditions, highly variable, and difficult to predict; and (3) occurs within a complex adaptive system with multiple components, where results are often greater than the sum of their parts.